Design and Evaluation of 3D-Printed Lattice Structures as High Flow Rate Aerosol Filters Authors: Yinkui Yu, Ning Zhang, Dominic Hoffman, Dewansh Rastogi, Ian R. Woodward, Catherine A. Fromen Published: 2024 ABSTRACT Aerosol contamination presents significant challenges across various industries, ranging from healthcare to manufacturing. Over the past few years, open foam filters have gained prominence for their ability to efficiently capture particles while allowing reasonable airflow. In this work, we present the use of 3D-printed idealized open foam-like lattice structures as aerosol filtration media, leveraging advances in additive manufacturing to generate these highly tunable and modular filters. Using parametric design approaches, we fabricated lattice filters with four different unit cell geometries (Cubic, Kelvin, Octahedron, and Weaire–Phelan) via Digital Light Synthesis 3D printing and characterized these structures with X-ray microcomputed tomography. We compared the aerosol filtration performance of the different lattice unit cell geometries using 1 μm polystyrene latex (PSL) aerosol particles, finding the filtration performance to be positively correlated with the single-unit-cell specific surface area. We then expanded our evaluation of deposition efficiency in Kelvin cell lattice structures of varied porosities, again finding a correlation between the specific surface area and deposition performance. Experimental analysis confirmed that deposition primarily occurs through impaction and electrostatic mechanisms within the parameter space. Overall, our findings demonstrate that unit-cell-based lattices can achieve a wide range of aerosol filtration efficiencies (∼10–100%) across various operating conditions (1–4 m/s superficial velocity), offering a highly tunable in-line filtration medium capable of maintaining high efficiency even at elevated airflow rates. This work not only provides essential guidelines for designing and manufacturing 3D-printed lattices as customizable aerosol filters but also highlights the current limitations and challenges in producing these structures. Introduction The history of aerosol filtration can be traced back to the early recognition of airborne disease transmission.1 Since then, various filtration techniques have emerged, each targeting specific aerosol particle sizes and types. Among these techniques, open foam filters have found extensive application due to their low pressure drops.2,3 These filters have been adopted across many industries, from healthcare facilities to industrial settings, largely to maintain air or water quality and remove particulates. Plastic and metal open foams have been proven to have good performance in filtering aerosol particles, but the tunability of their structural and material properties is still limited by polymerization techniques and templating methods. For example, the reticulated polyurethane foams tested by Clark et al. were made by removing cell membranes to leave a skeletal structure through chemical etching or thermal reticulation,4 and metal foams tested by Hellmann et al. were made by replicating the structure of polymer foams by investment casting or metallic deposition.2 Notably, none of these conventional methods are able to control or fine-tune the local structure of the foam cells during the manufacturing stage, resulting in largely stochastic foams.5 Thus, material integrity can be challenged when the fluid motion becomes chaotic, limiting the throughput of filtration processes and their widespread implementation.6 The advent of 3D printing technology has revolutionized traditional manufacturing processes and offers new untapped opportunities to provide highly tunable and controlled materials toward aerosol filtration. In general, 3D printing enables manufacturing of complex geometries with high precision and has found applications in diverse fields, including aerospace, medicine, and architecture.7 The unique advantages of 3D printing lie in its rapid prototyping capabilities and in enabling the creation of intricate structures that were previously unattainable through conventional methods. More recently, 3D printing has found a home beyond prototyping in the production of end-user consumer parts, highlighting the increasing maturity of the field.8 In the context of aerosol filtration, there have been notable recent examples utilizing 3D printing technologies, including 3D-printed examples such as biomimetic-villus for dust filtration, particle filters inspired by animal noses, downstream support structures for oil mist filters, and nanofiber membrane filters.9−12 Such recent advances in 3D printing have enabled the manufacturing of lattice structures as a promising alternative to traditional open foam filters. Lattices are one class of additive-only structures that have great potential in directing transport phenomena because they are highly ordered, scalable, and modular.13 Similar to foams, lattices are high-porosity, strut-based structures that are defined by repeating unit cells.14 Unlike their stochastic counterparts, lattices can be designed with defined dimensions, porosity, and surface area in an ordered arrangement, enabling a plethora of opportunities for various fluid flow applications. With an almost limitless design space, new structures and an extensive range of length scales can be designed and exploited for well-defined flow profiles to promote particle mixing or separation. Recent work from our group and others have demonstrated an ability to manufacture and evaluate single-phase airflows through sizable lattice structures, finding that they behave like a packed bed with overall low pressure drops.15−18 Thus, by virtue of their intricate and regular geometry, lattice structures may simultaneously facilitate efficient aerosol collection while maintaining desirable airflow rates under transitional to turbulent flow regimes, thereby addressing the long-standing trade-offs faced by conventional filters. In this article, we delve into the design space of 3D-printed lattice structures for aerosol filtration. Through a series of investigations, we fabricate lattices with varied unit cell geometries, porosity, and surface area to evaluate how these tunable material properties impact their filtration as a function of different aerosol aerodynamic particle diameters (dae) and interstitial flow velocities (Ui). Through this exploration, we aim to contribute to the advancement of lattice structures toward a wide range of aerosol filtration applications. Materials and Methods Lattice Selection and Design In this work, we investigated four different lattice unit cell geometries (shown in Figure 1a) including Cubic, Kelvin, Octahedron (or body-centered cubic with XY connections/BCCxy), and Weaire–Phelan (WP). These unit cell geometries were selected because of their distinctive macroscopic material properties, including porosity (ε) and specific surface area (av) under the same design parameters, unit cell size (lc), and dimensionless strut radius (r*). All unit cells were printed at a common lc = 2.38 mm at their maximum av, corresponding to variable r* values of r*cubic = 0.275, r*Kelvin = 0.15, r*octahedron = 0.15, and r*WP = 0.10. Single-cell-level theoretical ε and av as calculated from in silico CAD designs for a range of dimensionless strut radius (r*) are shown in Figure 1b,c for the four unit cell geometries studied in this work. (a) Visualizations of four different single unit cell lattice structures with unit cell closed mesh (r* = 0.1) and unit cell skeleton (inset, upper right). Strut radius, r, and unit cell length, lc, are labeled. (b) Theoretical single unit cell porosity (ε) of four different lattice geometries in parametric design as a function of dimensionless radius (r*). (c) Theoretical single unit cell dimensionless specific surface area (av × lc) of four different lattice geometries in parametric design as a function of dimensionless radius (r*). We also designed a series of Kelvin unit cells to investigate how the aerosol filtration performance shifts with ε and av. We selected four distinct r* values (r* = 0.075, 0.1, 0.125, 0.15) that fall within a suitable printability window while also enabling the investigation of different ε and av values. To improve printability of low r* lattice designs (r* = 0.075) and to prevent clogging from the additive manufacturing process, lc was increased from 2.38 to 3 mm.16 In all of our designs, unit cells with specified geometries and parameters homogeneously populated a cylindrical column primitive with a diameter of 12 mm and a height of 10 mm. The primitive structure of all lattices was designed to be a cylindrical column shape with fully conformal skin to enable fitting in a pipe while increasing the structural stability during printing and under high airflow rates.16 Lattice Manufacturing and Characterization The lattice design and generation approach was adopted from a published lattice generation method designed by our group.16 In our approach, the lattice generation process has four major steps: voxelization, population, inclusion and intersection, and meshing. Details and implementation of the generation procedure can be found in the prior work.16 We designed the varied lattice structures using Rhinoceros 7 and Grasshopper (Robert McNeel & Associates) on a desktop computer running an Intel E-2126G processor with 64 GB RAM. All lattices were manufactured with a Carbon M1 platform, using UMA90 resin (Carbon, Inc.), to achieve high tunability and fast prototyping speed in the design and manufacturing stages. The prints were set to use standard slicing with a z-axis resolution of 100 μm and x- and y-axis resolution of 75 μm. Following prints, all lattice structures were washed with isopropanol for 5 min in a Branson 1800 sonicator to remove excess resin and mitigate pore-clogging. However, through multiple stages from lattice design to postprocessing of 3D printing, undesirable defects can be introduced, including pore-clogging and strut-thinning.16,19,20 To characterize and study deviations between designed and printed macroscopic properties, namely, ε and av, the printed lattices were characterized by X-ray microcomputed tomography (micro-CT) with a Brüker Skyscan 1276 with a resolution of 20 μm. The scans were reconstructed using 3D slicer v5.0.3.21 A threshold filter using the Otsu algorithm was applied to slices to remove scan noise and reduce reconstruction artifacts. 3D models were generated with default smoothing settings (0.5 smoothing and 0 decimation), and then the volume and total surface area (a) were calculated with a segmentation statistics tool built in the 3D slicer. Notably, av, a single unit cell property, was used for designing single unit cells that populate the full lattice column, while a, a property for an entire lattice part, was used to evaluate the surface area of the physical printed lattices with conformal skins. From micro-CT reconstruction, the CT-av and CT-volume were determined for the as-printed parts, and a corrected r* (rc*) could be determined. Using rc*, CT-corrected values for av and volume were then calculated. Aerosol Generation and Deposition Measurements Our experimental flow system is shown in Figure 2a. The 3D-printed lattice column under investigation was mounted in a vertical pipe with a length of 0.79 m, with two sampling ports located upstream and downstream of the lattice sample. Aerosol deposition efficiency calculation details are shown in Supporting Information, Section 2, eq S1. The inner diameter of the sampling port was designed to be 3.5 mm. The flow was generated by a Copley HCP5 vacuum pump and regulated by a Copley TPK 2000 critical flow controller and a Copley DFM 2000 flow meter. Our operating flow rate range was 4–30 L/min, corresponding to lattice interstitial velocities ranging from 1 to 4 m/s. Interstitial velocity (Ui), the average velocity within the lattice, is defined as the superficial velocity (Uf), the bulk flow prior to interacting with the lattice, normalized by material porosity (ε) as shown in eq 11 (a) Experimental setup used for measuring aerosol deposition in lattices. (Neutralizer is a detachable component in the setup.) (b) Experimental setup used for pressure drop measurements. A TSI 3012A Kr-85 neutralizer was introduced upstream of the lattice column and used to remove electrostatic charges on aerosol particles when the contribution of electrostatic attraction deposition. Electrostatic attraction is reported as an important deposition mechanism especially for PSL particles for face masks by Laengert et al.;22 therefore, we also investigate the filtration performance difference with and without electrostatic effects by using a TSI 3012A Kr-85 neutralizer. Kr-85 is a radioactive β emitter, which creates ionized gas molecules within the neutralizer’s chamber. As PSL aerosol particles travel through the chamber, charged aerosol particles will attract ionized gas molecules with opposite charges to achieve charge neutralization. The neutralizer is graded to effectively neutralize up to a flow rate of 50 LPM that is suitable for our study (Umax = 34 LPM). To generate aerosol particles of controlled sizes, we suspended monodisperse polystyrene latex sphere (PSL) microbeads (Polyscience Polybeads Microspheres) of five different sizes, including 1.06, 2.07, 3.1, 4.78, and 5.93 μm, in methanol and aerosols were generated with a Collison jet nebulizer (CJN). The emitted aerosol concentration was measured by a TSI 3330 Optical Particle Sizer (OPS) with five channels. By controlling the suspension concentration and inlet flow rate of CJN, we controlled the aerosol concentration to be less than 3000 #/cm3 to reduce the coincidence error of the OPS and to reduce particle aggregation. Furthermore, by only adding monodisperse particles with specific sizes, the machine counting error resulting from miscounting across bins can be reduced. PSL particles are also considered nonhygroscopic particles that do not grow with increasing humidity. Based on our aerosol material selection, we decided not to include a diffusional dryer in our experimental setup. Prior work from our group has estimated the evaporation time of methanol around the polystyrene droplets to be sufficiently rapid, allowing us to consider only the PSL particles as monodisperse aerosols.23 For all deposition experiments, the inlet airflow rate of the CJN was set to 10 L/min, and a 3D-printed open connection was used to allow ambient air to flow into the column under higher airflow rates. Lattice samples were exposed to aerosols under controlled flow conditions for 5 min, and deposition was calculated by alternatively measuring particle concentrations upstream and downstream of the lattice filter via the sampling ports with the OPS. Each measuring window was 30 s, and there were five upstream and five downstream measurements in the 5 min exposure time. Deposition measurements are reported as the averaged value of 5–15 independent measurements taken across 3 different days. One-way ANOVA tests were performed via the Python 3 Scipy library, comparing interstitial velocity, dimensionless strut radius, and particle diameters to deposition efficiency. A p-value of 0.05 was selected to be of standard significance. Pressure Drop Measurements and Modeling Pressure drop was measured for all Kelvin cell lattices in a pipe setup, as shown in Figure 2b. The pressure measurements were taken in the same pipe used for aerosol deposition tests but were modified to match the experimental methods used in previous studies for cubic lattice and foams.15,17 Pressure gauges were placed upstream and downstream of the lattice sample region. Three different Dwyer Magnehelic differential pressure gauges were used to span the range of pressure drops tested: (1) 0–248.84 Pa, ±2%, (2) 0–746.52 Pa, ±4%, and (3) 0–2 kPa, ±5%. During the measurements, the flow rates were manually set on the vacuum pump as measured with a TPK 2000 to match a superficial velocity of 0–4 m/s with step increments or decrements of 0.2 m/s. For each lattice, two independent replicates of measurement were taken: one with step increments and the other with step decrements. Two replicates were combined to plot the pressure drop trends. Previous work in our group and others suggests that Kelvin cell lattice columns behave similarly to packed bed columns.15,18 Thus, pressure drop ΔP across homogeneous lattice columns with thickness t can be modeled by viscous and inertial contributions using the following equation:2where ρ is the fluid density, η is the fluid viscosity, and τ is the tortuosity of the lattice unit cell. The tortuosity of Kelvin cell lattices was calculated by an equation proposed by Inayat et al.,18 and details are shown in Supporting Information, Section 4, eq S8. Finally, dh is the hydraulic diameter of the lattice structure, which is defined as243where Ac in eq 3 is the cross-sectional area and pwetted is the wetted perimeter. In this study, we are operating at relatively high velocities (1–4 m/s), and air is considered a low-viscosity fluid; therefore, we chose to neglect the viscous term and keep only the inertial term. The packed bed Reynolds numbers for Kelvin cell lattice structures under operating flow conditions are calculated using eqs S9 and S10 and reported in Table S.14. The lowest packed bed Reynolds number starts at 50 and increases above 100, which justifies the choice of dropping the viscous term for simplicity of modeling. To further account for different strut radius thicknesses in Kelvin cell lattice studies, we introduced a geometrical compensation factor B that can be fitted from our data.25B is used to compensate for the difference between estimated properties, including hydraulic diameter and tortuosity, and physical properties. Equation 2 is truncated to4 Results Lattice Designs and Printability There are several types of additively manufactured lattice structures that have been created largely for part light-weighting, including strut-based and surface-based lattices.26 In this study, we selected four strut-based lattice unit cell geometries (Figure 1a) and evaluated their aerosol filtration performance. However, prior to fabrication, we had to first select specific design parameters to compare across different unit cells. In cellular lattice structures, the design flexibility is vast, but the structures are usually specified with a small number of parameters for convenience. Unit cell geometry, unit cell size (lc), strut radius (r), and primitive shape, that is, the overall shape of the entire latticed structure, are commonly adopted to achieve specified structures and functions. However, these design characteristics are not independent. Previous work in our group has shown that the cell length (lc) naturally emerges as a reference length scale for comparing features across lattice geometry and revealing self-similarities; furthermore, a dimensionless strut radius (r* = r/lc) can be used to predict theoretical quantities across different unit cell lengths and strut thicknesses for scaling purposes.15 For all unit cells, as r* increases, ε of a lattice unit cell decreases monotonically because the solid struts gradually take on more volume (Figure 1b). In contrast, av increases as r* increases until a threshold r* (rt*) is exceeded, and then av decreases (Figure 1c). At r* values above the threshold, the pores decrease in diameter, and the exposed surface area is thus reduced. We call this phenomenon “overcrowding,” and this design region is usually not preferred because of (1) an increased pressure drop without increasing surface area, (2) merging during printing resulting in loss of local geometrical features, and (3) low ε and potential clogging in postprocessing. As shown in Figure 1c, the rt* (i.e., r* of maximum av) values are 0.275, 0.15, 0.15, and 0.1 for Cubic, Kelvin, Octahedron, and WP, respectively. Interestingly, these rt* values correspond to a similar single unit cell porosity around 0.57 (between 0.53 and 0.61), making this a unique comparative point to hold ε and lc similar while varying r* and av. For aerosol evaluations of each unit cell type, we selected the rt* at the peak av and adopted an lc of 2.38 mm from our previous work.15,16 These design parameters allowed us to compare aerosol filtration performance under similar porosities but different av. We also further studied the aerosol filtration performance of Kelvin cell lattices to probe the role of ε and av on aerosol filtration. We selected Kelvin cells for this in-depth analysis since they are commonly regarded as an idealization of traditional stochastic open foams due to their similarities in structural, thermal, and mass transport properties.15,27,28 A Kelvin unit cell has three different pore dimensions as shown in Figure S.3 and can be calculated by eq S5, leading to an overall tortuosity that largely prevents a straight path through the sample.18 Choosing a unit cell length lc of 3 mm instead of 2.38 mm, to ensure printability of thin strut lattice (i.e., r* = 0.075), for all the lattice parts sets the pore diameters between 0.45 to 2.76 mm. Of note, these pores are approximately 1–2 orders of magnitude larger than traditional polyurethane or nickel foams.2,4 Following these design choices, we were able to successfully create different lattices with varied unit cell geometries and properties. The digital designs of four full lattice columns lc = 2.38 mm and unit cell geometries of Cubic (r* = 0.275), Kelvin (r* = 0.15), Octahedron (r* = 0.15), and WP (r* = 0.1) are shown in Figure 3a, and pictures of the same parts following printing are shown in Figure 3b. For these lattices, CT-reconstructed results indicated different levels of pore-clogging, resulting in reduced porosity, as shown in Figure 4. Pore-clogging was a prominent issue for Cubic cell lattices and WP cell lattices, leading to a decrease in porosity values by about 12 and 15%, respectively. The pore-clogging of the Cubic cell lattice was related to the thick boundary supporting beam thickness, and the pore-clogging of the WP cell lattice was attributed to the small pore sizes of WP unit cell geometry. (a) Maximum specific surface area lattice columns with conformal skin support in CAD (top row) and physical 3D-printed samples (bottom row). Samples from left to right have geometries of cubic, Kelvin, octahedron, and Weaire–Phelan, respectively, with dimensionless strut radius (r*) labeled. (b) Kelvin unit cell series with conformal skin support in CAD (top row) and physical 3D-printed samples (bottom row), with dimensionless strut radius (r*) labeled. CT-reconstructed lattice sample porosity (red bar) and designed lattice sample porosity (blue bar) of Cubic, Kelvin, Octahedron, and Weaire–Phelan unit cell geometries. For Kelvin cell lattices, the digital designs of four full lattice columns of lc = 3 mm and designed r* = 0.075, 0.1, 0.125, and 0.15 are shown in Figure 3c, with pictures of the same parts following printing shown in Figure 3d. Through test prints, we also confirmed printability and pore-clogging region for the Kelvin cell lattices region, as shown in Figure 5. To compensate for the strut-thinning, rc* was measured by CT-reconstructed volume for Kelvin lattices (representative reconstructed parts are shown in Figures S.4–S.11). The ε and av of the lattice samples were obtained using eqs S6 and S7 and compared with designed values and corrected values, as shown in Table 1. In our study, the rc* values for 4 Kelvin unit cells were 0.060, 0.090, 0.116, and 0.143, respectively. The av after correction did not exactly match the CT-reconstructed values because (1) the strut-thinning is not globally uniform and (2) CT-reconstructed av values are usually biased by the specific smoothing method. Admitting these limitations, the corrected av and rc* can still be restricted within 10% deviation of the designed values and more accurately represent the lattice part used during tests. Theoretical single unit cell porosity (blue dashed line) and dimensionless specific surface area (red dashed line) of a Kelvin lattice in parametric design as a function of dimensionless radius (r*). Dimensionless specific surface area is defined as specific surface area av [mm–1] multiplied by unit cell length lc [mm] and is used to unify values for lattice unit cells with different lc. The shaded area indicates the printability (purple = nonprintable region, green = region used in this study, yellow = “overcrowded” region). Table 1Comparison of Kelvin Cell Lattice Column Volume and Surface Area (av) among the Designed, CT-Reconstructed, and Corrected Valuesadesigned r*CT-rc*designed volume [mm3]CT-volume [mm3]CT-corrected volume [mm3]error of volume after correction [±%]designed a [mm2]CT-a [mm2]CT-corrected a [mm2]error of a after correction [±%]0.0750.060168112 ± 3111.70.2614881198 ± 211296–7.540.10.090278236 ± 4236.0017171545 ± 101651–6.600.1250.116402362 ± 2361.20.2218041669 ± 31794–6.970.150.143521489 ± 5489.9–0.1817871674 ± 191797–6.85aReported values are the mean and standard deviation from three replicate parts. Comparison of Kelvin Cell Lattice Column Volume and Surface Area (av) among the Designed, CT-Reconstructed, and Corrected Valuesa Reported values are the mean and standard deviation from three replicate parts. Aerosol Filtration and Deposition Mechanisms Using the printed lattice columns of varied geometries, we first evaluated deposition efficiency (F) for PSL with a dae of 1.06 μm in the described Cubic, Kelvin, Octahedron, and Weaire–Phelan cell geometries under three different Uf of 1, 3, and 5 m/s. Results are reported in Table S.11 and shown in Figure 6a. At a Uf of 1 m/s, we observed the highest F in the WP lattice (83%) and the lowest F in the Cubic lattice (16%). Kelvin and Octahedron had similar F efficiencies of 48.9 and 38.7%, respectively. As the Uf increased to 3 and 5 m/s, the difference in F between geometries decreased and eventually reached F values larger than 95%. We then converted the face velocity Uf into interstitial velocity Ui and replotted the data in Figure 6b. We observed F correlated with av, where unit cell geometries with higher av have better filtration performance that is most apparent when plotted against Ui, a unit cell property variable between unit cells, instead of Uf, which was constant across samples. These studies highlight that av is a significant determinant of F under these similar homogeneous unit-cell-based lattice designs, which is a property that can be tuned relatively independently from ε, which is the critical determinant in the overall ΔP across the filter. (a) Deposition efficiency (F) with respect to superficial velocity (Uf) for lattice columns with four different lattice unit cell geometries. (b) F with respect to interstitial velocity (Ui) for lattice columns with four different lattice unit cell geometries. Error bars represent mean ± one standard deviation of the deposition efficiency (n = 3). Aerosol particles used in this study are monodisperse polystyrene latex particles (PSL) with an aerodynamic diameter (dae) of 1 μm. (c) Flow direction relative to unit cell orientations. To confirm these findings, we next evaluated F in Kelvin cell lattice columns with four different dimensionless radius rc* under a range of flow rates and dae. The series rc* of 0.060, 0.090, 0.116, and 0.143 yielded parts with corrected av values of 1296, 1651, 1794, and 1797 mm2 (Table 1), allowing evaluation at a range of av. Results from studies for PSL of varied dae (1.06, 2.07, 3.1, 4.78, and 5.93 μm spheres) are reported in Tables S.6–S.9 and shown in Figure 7. One-way ANOVA tests (results shown in Table S.1) indicated that dae (p < 0.001), Ui (p < 0.001), and rc* (p < 0.001) were statistically significant factors of F. For filters of a given rc*, under the same Ui, F increased as dae increased. Additionally, for the same dae, F increased as Ui increased. Cross-comparing Kelvin cell lattice column with different rc*, we observed F improvement as the rc* and correspondingly the av increased. Similar to traditional foam filters, the F improvement for smaller particles was less significant compared to larger particles across all lattices.4 As with open foam filters, the Kelvin cell lattices more efficiently capture aerosol particles with larger dae due to their greater inertia. For a particle size of 1 μm, F ranged approximately from 5 to 70% depending on Ui and rc*. In contrast, the F of 3 μm particles ranged approximately from 25 to 100% and the F of 6 μm particles ranged approximately from 70 to 100%. Both Ui and dae are positively correlated with the Stokes number, a nondimensional number that characterizes deposition via impaction. In Table S.12, quantitative characterizations of nondimensional numbers using eqs S2–S4 of different deposition mechanisms suggest that impaction is the major deposition mechanism under the parameters investigated here.29 Estimations of gravitational settling and diffusive transport contributions were insignificant compared to the impaction contribution.25 Deposition efficiency (F) with respect to polystyrene aerosol aerodynamic diameter (dae) for Kelvin lattice columns with different dimensionless radii (rc*) under different interstitial velocities (Ui). Panels (a) to (d) show the data from samples with rc* of 0.060, 0.090, 0.116, and 0.143, respectively. Points show the mean of 10–15 individual experimental measurements. Error bars represent ±1 standard deviation of the deposition efficiency (n > 10). Another important deposition mechanism in aerosol filtration is electrostatic attraction.30 We chose a Kelvin cell lattice sample with a r* of 0.15 (rc* = 0.143) at an Ui of 4 m/s to test the electrostatic contributions in our system. Results from this study are reported in Table S.10 and shown in Figure 8. One-way ANOVA tests indicated that the use of a neutralizer, which mitigates electrostatic changes in the PSL aerosols, is a statistically significant factor of F (p < 0.05). Through the addition of a neutralizer, F remains positively correlated with dae, pointing again to the role of impaction as a dominating deposition mechanism. However, the deposition efficiency decreases by 20–44% at different dae from the charged PSL spheres. These results show that electrostatic charge effects also play an important role in aerosol deposition in the lattice structures. This effect may be amplified because of the use of PSL aerosol particles22 and polymer resin as lattice printing materials,31 where both plastic materials are known to harbor electrostatic charges. Due to the flexibility of material selection in additive manufactured lattice structures, this study also suggests it might be possible to boost the electrostatic charge effects and improve the filtration performance of charged particles by using conductive materials to make the lattices and by supplying a charge to the lattice structure. Deposition efficiency (F) with respect to polystyrene aerosol aerodynamic diameter (dae) for the Kelvin lattice column (r* = 0.15, Ui = 4 m/s). Round points represent deposition measurements without using a charge neutralizer and square points represent deposition measurements with using a charge neutralizer. Points show the mean of five individual experimental measurements. Error bars represent ±1 standard deviation of the deposition efficiency (n = 5). Log scale (with a base of 10) deposition efficiency values for charge particles with sizes of 3.1, 4.78, and 5.93 μm through Kelvin lattice column (r* = 0.15, Ui = 4 m/s) can be found in Figure S.1. Kelvin Cell Pressure Drop Measurements To evaluate their real-world applicability for aerosol filtration studies, we performed a series of pressure drop measurements within the four Kelvin cells to establish baseline flow properties, in line with our prior work. The pressure drops of the lattice with different rc* as a function of superficial velocity are shown in Figure 9. The experimental pressure drop measurements are reported in Tables S.2–S.5. We find that pressure drop increases as rc* increases; the truncated pressure drop model (eq 4) in the dashed line captures the pressure drop trend well. (Values of geometrical coefficient, B, and other calculation details can be found in Table S.13.) B increases as porosity decreases, and this suggests that the estimated tortuosity values using eq S8 may be smaller than the actual tortuosity value. The pressure drop of rc* = 0.143 (green) is almost 2-fold the pressure drop of rc* = 0.116 (magenta); however, this does not correspond directly to a 2-fold increase in deposition, as shown in Figure 7. Overall, these results suggest that the lattice filter can offer high flow throughput aerosol filtration with a low to moderate pressure drop. Pressure drop of Kelvin lattice columns with different rc*. Dashed line indicates a quadratic fit with first and zero order coefficients fixed to 0. Discussion In our studies, lattice structures demonstrate remarkable tunability, enabling a broad range of deposition efficiencies from 5 to 100% for aerosol particles within the 1–6 μm size range. We find that lattice unit cell geometry affects the filtration performance of lattice filters, and generally, lattice structures with higher surface area (controlled by av at the unit cell scale) have better filtration performance. Kelvin cell lattices especially behave similarly to traditional open foams but can be operated at Ui up to 4 m/s with only moderate pressure drops to enable high flow rate filtration. These lattices show comparable particle filtration efficiency to open foam filters at variable operating conditions.2 We also investigated and confirmed inertial impaction and electrostatic attraction as major deposition mechanisms for the particle size and flow rate ranges that we studied. These studies merely scratch the surface of the potential for lattice structures as filtration media, laying a crucial foundation for systematically evaluating key design choices needed to enhance the filter performance. As 3D printing techniques continue to advance, exploring diverse unit cell configurations and reducing dimensions will further improve the filtration efficiency while minimizing pressure drops. The flexibility of 3D printing allows for a swift transition from design to implementation, meeting customized filtration needs optimized for diverse applications. We anticipate growing design flexibility for lattice structures, opening up diverse applications, such as industrial and environmental aerosol filtration, inhalable drug deposition evaluations, and rapid prototyping of filters. While the included studies demonstrate a vast potential for lattices in aerosol filtration, there are a few limitations to our studies that require additional investigations. Notably, the limited size range of particles investigated in this work spans only from 1 to 6 μm. This constraint arises due to our approach of mixing monodisperse particles to achieve more accurate measurements within the optimal operating OPS size range of 0.5–10 μm. Regarding the flow velocity, we selected interstitial velocities between 1 and 4 m/s to optimize sampling conditions.25 Challenges are also presented by the OPS sensitivity, primarily due to the difficulty in controlling the error margins.32,33 This uncertainty limits the choice of evaluating Kelvin cell lattices with finer strut radius variations in the tested range, making it difficult to obtain results that are statistically significant. Also, our primitive shape of lattice column has a relatively small size (12 mm in diameter and 10 mm in thickness) compared to unit cell length (2.38 and 3 mm). In this case, the dimensions of the primitive shape can only span 4–6 unit cells, where entrance flow effects may play a significant role. As shown in Table S.14, packed bed Reynolds numbers of the Kelvin cell lattice under our operating flow conditions indicate that the systems are in a transitional flow regime, and therefore, our deposition efficiency measurements may be biased to some extent by entrance and wall effects. While we do not directly assess these effects in our current work, we do find a linear correlation between ln(1 – F) and the lattice thickness (Figure S.2), as would be suggested by models of traditional open foams;4 this trend suggests that we are largely observing bulk behavior even in these relatively small lattice structures although more work is needed in this regard. Certainly, there exists a vast design space for lattice unit geometry other than the Kelvin, Cubic, Octahedron, and Weire–Phalen demonstrated in this work,16,26 and we plan to extend our investigations of aerosol deposition to a wide range of lattice geometries in the future. Beyond the above limitations of this work, we wish to note three major challenges remaining in the broader use of lattices in aerosol filtration applications: (1) manufacturing of lattices with less error from the designed properties, (2) quantitively modeling the aerosol deposition efficiency across different lattice geometries and design parameters, and (3) modeling the intricate flow pattern within the homogeneous lattice structures to further predict aerosol behavior.2−4 The first challenge will require advancement in 3D printing technologies and improved design-stage strategies for process compensation, which will vary with the overall manufacturing process.16,26 The second challenge is hard to resolve by previous semiempirical modeling approaches since the design freedom of lattice filters is vast and tunable. A potential solution lies with the abilities of machine learning technology; after the collection of a sufficient amount of training data, a higher dimensional model consisting of multiple design parameters can be created and trained. The third challenge can be addressed by computational tools such as computational fluid particle dynamics,15,34−36 enabling the potential of fully digitalized modeling from design to applications. However, due to the large size of the overall lattice primitive shape, intricate coupled boundary geometries, and high operating flow velocities, additional research is necessary to use such computational tools to conquer these challenges. Despite these limitations and challenges, 3D-printed lattice structures represent an entirely unique filtration material that may enable vast new applications. Compared to open foams, lattice structures have pore sizes that can be 1–2 orders of magnitude larger, which naturally promotes higher filtration throughput and reduces resistance through the structure.2,4 Open foams need to be characterized after manufacturing to determine overall material properties and are highly sensitive to manufacturing processes; further, the complex manufacturing process creates inherent variabilities. The stochastic nature of open foam formation results in filters with a wide range of properties produced from the same manufacturing process. In contrast, lattice structures have controllable material properties that are highly reproducible, tunable, and homogeneous. This minimizes the time requirements from design to manufacturing for new applications. Conclusions In this study, the efficacy of using 3D-printed lattice structures as innovative aerosol filtration media has been examined. Cylindrical lattice filters were additively manufactured and characterized to obtain material properties, including pore size, surface area, and porosity. Experimental aerosol deposition results show that lattice unit cell geometries with a high surface area have better deposition performance. Additionally, the results indicate that Kelvin cell lattices can achieve a wide range of aerosol filtration efficiencies with great tunability, even at relatively high flow rates. Inertial impaction and electrostatic attractions are identified as the major deposition mechanisms in the investigated flow range of 1–4 m/s and particle diameters over 1 μm. To our knowledge, this represents some of the first studies into 3D-printed lattice structures for aerosol filtration, building a critical foundation for future efforts in this high-potential area. Overall, our work demonstrates the utility of 3D-printed lattice structures as an alternative to stochastic open foams that show tremendous potential as new candidates for aerosol filtration in multiple applications, including industrial manufacturing, healthcare, medical applications, and personal protective equipment.